You are on page 1of 8

2010 International Conference on Power System Technology

Vehicle to Grid - Monte Carlo simulations for


optimal Aggregator Strategies
Claes Sandels, Ulrik Franke, Student Member, IEEE, Niklas Ingvar, Lars Nordstrom, Member, IEEE,
Roberth Hamren

Abstract-Previous work has shown that it could be profitable related to their availability [6], and compared to conventional
on some control markets to use Plug-in Hybrid Electric Vehicles control power providers they can provide the power extremely
(PHEV) as control power resources. This concept, where battery
fast [7]. The concept of electric vehicles that are connected
driven vehicles such as PHEV s provide ancillary service to the
grid is commonly referred to as Vehicle to Grid (V2G). The
to the grid and can retrieve and inject controlled amounts
idea is to sell the capacity and energy of the parked PHEVs on of electric energy from/to the grid is often referred to as
the control market. Due to the fact that cars on average are Vehicle to Grid (V2G). Deeper studies of the V2G concept
parked 92% of the day, the availability of this capacity could be can be found in [3], [7], [8]. Several small scale tests of this
very high, even though it will be highly dependent on commuting
concept have been successfully conducted [9]. However, the
patterns in peak hours. However, as each PHEV has a very small
capacity from a grid perspective, it is necessary to implement an
challenge that lies ahead is to study the technical feasibliliy of
aggregating control system, managing a large number of vehicles. having thousands to hundreds of thousands aggregated PHEVs
This paper presents strategies for an Aggregator to fulfill control providing this control power. This must be done fulfilling the
bids on the German control markets. These strategies are tested requirements of an available, reliable and secure power system
with respect to reliability, efficiency and profitability in a Monte
for the system operators.
Carlo simulation model. The model is based on available data
on the distributions of commuting departure times and travel
distances, as well as average driving power consumption, PHEV A. Scope of the paper
battery capacities and the market constraints of the secondary
control market in Germany. The main scope of this paper is to derive strategies for an
Index Terms-Vehicle to Grid, Control Market, Aggregator, Aggregator to act on the German control markets, i.e. how to
Plug-in Hybrid Electric Vehicles, Monte Carlo simulations control the PHEVs in order to fulfill the bids of the Aggregator.
After this, an evaluation of these processes are performed
in a Monte Carlo simulation model. The model is based on
I. INTRODUCTION
available data on the distributions of commuting departure
Predictions suggest that there can be as much as 6.7 million times and travel distances, as well as average driving power
Plug-in Hybrid Electric Vehicles (PHEV) in Germany by the consumption and PHEV battery capacities. The simulations
year 2020 [1]. With the assumptions that all these PHEVs will try to answer the questions whether it is efficient, reliable
are connected to the grid and using a 11.1 kW connection, and economically profitable for PHEVs to provide control
the German PHEV fleet will have equivalent power to match power on today's markets in Germany. Note, that no technical
the short term capacity of 23 nuclear power plants [2]. Due constraints of V2G will be discussed in this paper.
to the fact that cars in average are parked 92% of the day
[3], the availability of this enormous capacity will be very
B. Outline
high. The government of Germany has declared that 40% of
the greenhouse gas emission have to be reduced until 2020. The remainder of this paper is structured as follows. Sec­
This has to be done from a starting-point where Germany is tion II contrasts the present contribution with some related
generating 82% of its energy from fossil fuels [4]. Due to the work in the fields of PHEVs and V2G. In section Ill, a back­
intermittent character of some renewable energy production ground of the German control markets is presented, including
such as wind generated energy, an increased demand for the legal responsibilities this implies. Here, the strategies for
control power is predicted. The idea is to sell the capacity an Aggregator are also presented. It is followed, in section IV,
and energy of the parked PHEVs on what is called the control by a Monte Carlo simulation model that allows the strategies
market [5]. The control market, in short, is a marketplace of the Aggregator to be evaluated in a quantitative manner.
where an actor can be ready to provide backup power (control Section V contains a discussion of the results, and some
power) and energy if something unpredicted would happen, concluding remarks are given in section VI.
e.g. a breakdown in a power plant or that the wind stops
blowing at the wind power farms. A competitive factor of II. R ELAT ED WOR K
using PHEVs on the control market is that there is no cost
In [10], the profitability and suitability of the participation
Sandels, Franke, Ingvar and Nordstrom are with the Department of Indus­ of electric vehicles (EVs) at the control markets in the US are
trial Information and Control Systems, Royal Institute of Technology, Stock­ studied. Subsequently, cost and revenue equations are derived
holm, Sweden, e-mail: c1aes.sandels@gmail.com. { ulrikf. larsn } @ics.kth.se.
niklas@ingvar.com. Harnren is with Vattenfall Research and Development, for these markets. By inputting power market data into the
Stockholm, Sweden, e-mail: roberth.harnren@vattenfall.com. model, the authors can conclude that the markets have the
978-J-4244-5940-7/1O/$26.00©201O IEEE
2

potential to be profitable for EVs. Since then, similar studies


have been composed for various control markets [II], [12],
[13].
The authors of [7] present a conceptual framework for a
V2G concept in the US. This framework defines how the EVs
are contracted by a central entity, called the Aggregator, to
service the power grid. Here, a lot of emphasis is put on
a proposed IT system to handle the defined control signals
and communication flows between the various actors, e.g. the Fig. 1. Interaction between the TSO, Aggregator and PHEVs.
PHEV owner, the Aggregator and the Transmission System
Operator (TSO). Finally, simulations are performed of an EV
fleet used for commuting purposes to study the control power time, based on human behavior, control market constraints are
capability. The simulations do not take any control market taken into consideration.
constraints into consideration, and, in addition, no economical This article aims to define strategies for the Aggregator, in
results are calculated. Instead, only the physical properties of order to be able to fulfill bids on the control markets. These
an aggregated EV fleet are presented, e.g. the stored battery strategies are backed up with simulations.
capacity and EV availability. The conclusion is that EVs have
great potential to provide control power. III. T HE GERMAN CONTROL MARKETS
Two different V2G architectures are derived for ancillary
In Germany, there are four profit driven TSOs responsible
services in the US in [14]. The first architecture proposes
for maintaining the balance between generation and consump­
a system where the TSO controls the plugged in PHEV di­
tion of electric energy in the power system. To maintain this
rectly. The second one proposes an Aggregator controlling the
balance, the TSO has three different types of control mech­
PHEVs. These two different architectures are later evaluated
anisms to utilize. These control mechanisms generate three
regarding availability, reliability and profitability. With the
different tendering control markets, where any actor fulfilling
input of, e.g. travel and regulation data into the model, it is
TSO requirements can bid. In Table I, these markets are
shown that the aggregated architecture is more reliable but
defined with respect to e.g. market constraints and properties,
less profitable than the other one. It is concluded that the
respectively. For more detailed information, see [18], [19].
aggregated architecture is more probable to be implemented
In the simulations described in section IV, we have chosen
in a near term ancillary services system.
to focus on the secondary control market. The reason is that it
A more detailed mathematical interpretation of the role of
has been shown to have the potential to be most profitable [6].
the Aggregator is found in [15]. With dynamic programming
In addition, asymmetric bids are feasible, which is considered
the authors maximize the customer benefit, i.e., the State Of
to be an interesting property for PHEVs. However, with the
Charge (SOC) before the next trip, and the total revenue of
results obtained from these simulations, a short discussion will
the Aggregator.
also take place regarding the suitability for acting on the other
In [16], simulations are performed regarding the charging of
two markets in section V.
the future EV fleet in Germany with respect to traffic behavior
based on data from the travel survey "Mobility in Germany"
[17]. The aim of this simulation is to, based on three different A. The secondary control market
charging strategies, study how this car fleet will affect the To understand the main principles of the simulations, some
power grid. The first conclusion of the simulations is that additional theory of the secondary control market is needed.
90% of the fleet is parked at any given time. Further on, it is Secondary control is needed for two reasons: to restore the
concluded that there can be a huge potential for V2G systems. primary control and to restore the frequency to its nominal
For example, the aggregated battery storage is significant, and, level of 50 Hz. In contrast to the primary control, the sec­
under conservative assumptions more than 1 GW of power is ondary control is activated with respect to where the actual
always available. deviation has occurred, i.e. the affected control zone and TSO.
Simulations with 500 participating PHEVs based on the Every actor that bids on this market defines bid sizes of control
control market constraints of Sweden and Germany are per­ power for peak hours (8 a.m. - 8 p.m. workdays) and/or off
formed in [6]. Here, it is shown that the German markets have peak hours (8 p.m. - 8 a.m. workdays and all holidays). Also,
the potential to be profitable (e.g. 82 €IPHEV/month on the energy prices for the capacities are set by the bidding actor.
secondary control market), whereas the Swedish markets have The actual activation of secondary control by the TSO is
no potential of being profitable in these simulations. then determined by these prices. Down control is activated
However, none of the reviewed articles present adequate by descending energy prices, and up control is activated by
description of how the Aggregator can act on the control ascending energy prices until the demand is met.
markets, i.e., aspects such as control market rules, customer
needs and technical battery constraints in order to fulfill
B. PHEVs participating on the control markets
the control bids. In addition, none of the articles perform
simulations of a PHEV fleet, where availability, reliability and Evidently, many constraints are set by the rules and regula­
the economical incentives of providing control power over tions of the TSOs. A mutual cross TSO rule is that there is a
TABLE I
THE DIFFERENT CONTROL MARKETS IN GERMANY [18] , [20] , [21] .
Type of control Activation Time requirements Bid requirements Procured Payments
Primary Automatic - frequency devia­ After 15 seconds 50% of the At least 5 MW symmetric, to Monthly Capacity. Pay as
tion (50±0.2 Hz). contracted power. After 30 be delivered around the clock bid.
seconds 100% of the con­ for one month.
tracted power.
Secondary Semi automatic - frequency Start activation after 30 At least 10 MW asymmetric. Monthly Capacity and en­
deviation but with respect to seconds. Complete activation Delivery is one month, and is ergy. Pay as bid.
affected control zone (50±0.2 within five minutes (at most). split into peak and off peak
Hz). hours.
Tertiary Manual - by telephone call Complete activation after 15 At least 15 MW asymmetric, Daily Capacity and en­
from affected TSO (control minutes (at most). to be delivered for four hours. ergy. Pay as bid.
zone).

FHB/s in standby for up


control. Active F+tEVs for up
,..---l.;____ ,.L, oontrol.
.,---:

9:)C>99%
Decrease load stop charging

}
+ =
Olarging PHEVsin
Increase Generation discharge battery
-' Part Generation
up
+ =
standby for control.
- Increase load start charging
/'O�
=

A PHEV can provide


control power S::X::> MinS::X:; D
....8
...
:Y......

}
( .. Up control) + Decrease load = stop charging
----J-._.L
(- Down control) - Increase load = start charging
Part Controllable Load Non charging A-tEVsin
standby for down
oon trol.
S::X:<
: Min9:X

Fig. 2. How PHEVs can provide control power.


Active PHEVs for down
oontrol. '0---0
-
F+tEVsin thisarea will PHBlsnot oonneded to grid
alwaysdlarge.
-�
'O---O! -
minimum bid capacity. Even the lowest bid (5 MW) is far too
large for a single PHEV to handle. Therefore, an aggregation Fig. 3. The grouping process performed by the Aggregator for part
Generation. Each area represents a set of PHEVs.
function needs to emerge, where many PHEVs are pooled as
one unit by an actor denoted as the Aggregator [7], [14]. From
the viewpoint of the TSO, the Aggregator will be one single
actor who delivers the promised control power. However, the • Information about connection capacity
physical delivery is done by each PHEY. If a symmetric bid by the Aggregator is won on the control
Any bidder on the German control markets must be a Bal­ market, his task is to ensure that promised control power is
ance Responsible Party (BRP), i.e. an actor who is financially available and can be delivered if needed. The responsibility to
responsible for maintaining balance between production and know which PHEVs that can, and will be deployed for control
consumption. This is not possible for a single PHEV owner. In power, is the central function of the Aggregator. Therefore, a
addition, the establishment of an Aggregator makes sure that process must be defined for how this can be accomplished.
the PHEV owner does not have to be engaged in administration This process will be denoted as the grouping process of the
processes, e.g. procurement, bidding and price settings with Aggregator.
the TSO. In Fig. 1, an illustration of the interaction between In Fig. 3, we can see the graphical representation of this
the TSO and the Aggregator is shown. grouping process for part Generation. With an aggregated
PHEVs can provide symmetric control power by just de­ PHEV fleet of size N, combined with data regarding battery
creasing and increasing the load on the grid, i.e. start to size, SOC, MinSOC, connection capacity and a triggered
charge a not charging PHEV to deliver down control, and signal as soon as a PHEV plugs in/plugs out, the Aggregator
stop the charging of an already charging PHEV to deliver can calculate the aggregated control power up and down, by
up control. But in addition to this, a PHEV can discharge placing the PHEVs into different sets. Note that the connection
their batteries and feed the electric energy into the grid as capacity is a function of the SOC when it comes to charging.
an extra feature to provide up control. This fact, results in A high SOC, implies less capacity, and this is denoted as the
two different configurations for delivering symmetric control actual transfer capacity ("'bat in Table II). The Aggregator then
power, see Fig. 2. These two configurations are denoted as makes sure that the up control bid (UC) fits within the sets A,
part Generation and part Controllable Load. Bl and B2, and, that the down control bid (DC) fits within B1.
When a customer enters the V2G service, certain informa­ The sizes of UC and DC vary. For example, if a PHEV owner
tion must be available for the Aggregator. This is required active in set Bl disconnects, and no new PHEVs connect, the
for the placement of correct control bids. Such information DC bubble will increase in size. However, if UC is smaller than
includes: DC, the Aggregator can move PHEVs located in B2 to B1,
• The battery capacity and battery type of the PHEV and increase the down control capability, i.e. decrease the DC
• Minimum Wanted SOC (MinSOC), see Table II bubble. Evidently, decisions made by the PHEV owner and the
4

Aggregator, will result in that the sizes UC and DC will vary 20


....----,,--,---i -&- Total ability 10 deliver down conlrol power
over time, and, his goal is to fit them within the respective sets - Delivered down control power

- The down control bid


1S
at all times, therefore, fulfilling his control bids to the TSO.
The graphical representation for the grouping process for part 10

Controllable Load is identical to the one for part Generation,


but, with the difference that the UC bubble only can fit within
B2.
time (hrsJ

IV. MONT E C ARLO SIMULATIONS 20 ..------,---�--_l


-- Delivered up control power

- The up control bid

To be able to study the properties of the suggested grouping



process of the Aggregator on the control markets, a math­ "
.� 10
ematical model is derived. This model will be used in a
1
limited scenario in Matlab. The purpose of the model is to
see how well the part Generation and the part Controllable
Load can provide control power, and to project the income timelhrs)

for them both, respectively. This will lay a foundation for a


comparison between the two parts. Later, a sensitivity analysis
Fig. 4. Ability to deliver control power for part Generation.
is performed in order to see how robust the two parts are with
regard to changes in parameter values.
The chosen scenario is to study PHEVs used for commuting 100

purposes. Commuting purposes imply that the PHEVs will be :>

e
driven twice a day, i.e. from home to work in the morning, '"
80
c

and from work to home in the afternoon/evening. �


:to
As mentioned in section III, these PHEVs will participate ij
'" 60
on the secondary control market. £
.�

isI 40
A. Data and assumptions 0..
"0
Evidently, in order to model a PHEV fleet and a control '"
'"
20

market, some data and assumptions are needed. c

In Table II, the parameters of the model are presented with �


0..

the corresponding numerical value and a comment. These 10 15 20


time [hrsl
parameter settings will be denoted as the Basic Case.
Fig. 5. Percentage of PHEVs in the different sets over the day for part
B. Results Basic Case Generation.

Here, the result and analysis of the performed Monte Carlo


simulations for the Basic Case will be presented for part
Generation and Controllable Load, respectively. places symmetrical bids, i.e. better up control power capability
In Fig 4, we can see the results from the simulations for part is obtained. This result is reasonable due to three reasons.
Generation displayed in two subplots. Each subplot displays I) The Aggregator can deliver up control power from three
two separate curves and one line. The circled curve in the sets, but only from one set when it comes to down control
upper subplot shows the aggregated down control power from power. 2) Transfer capacity is a function of SOC when it
the PHEVs over 24 hours. The crossed curve in the lower comes to charging a battery, while not when discharging. 3)
subplot shows the actual delivered control power by activation Down control is activated more often than up control by the
from the TSO. The straight line in the upper subplot states TSO, because of the bid structure of the competitors, and that
the down control bid of the Aggregator. The lower subplot is there is historically a bigger down control demand (see WTSO
composed in the same way, but for up control power instead. in Table II). Evidently, this results in frequent charging of
As we can see, the aggregated down control power is PHEVs in B1.
larger than the bid, except between the time 07:30 and 08:30. So, when down control has times when shortages arise, there
Therefore, a conclusion is that down control will be critical in is always a lot of surplus in up control power in comparison
the morning. This result is reasonable because most PHEVs to the bid. However, having a lot of surplus is not wise. If the
will arrive at home in the afternoon/evening and plug in their Aggregator has an up control bid of, e.g. 5 MW, but most of
vehicles. Due to the fact that down control is called more often the time there are at least 10 MW available from his PHEVs,
than up control, PHEVs in B1 and B2 will charge continuously then the potential up capacity payment for each PHEV will
and retrieve high SOCs. This SOC will not be reduced until be reduced by a factor two. Still, to ensure the down control
the PHEVs are driven to work in the morning. capability, this is a necessary decision by the Aggregator.
The aggregated up control power is always larger than In Fig. 5, we can see how the PHEVs move between the
the aggregated down control power, although the Aggregator different sets over the day. As seen, the amount of PHEVs
5

TABLE II
THE PARAMETERS IN THE MODEL AND CORRES PONDING BASIC CAS E VALUE.
Parameter Description Numerical value Comment
N Number of PHEVs 5000 PHEVs Considered to be a reasonable aggregation in a larger city of Germany.
MinSOC Minimum wanted SOC 20% of N have 30% MinSOC The MinSOC is the minimum level of the SOC that the PHEV
60% of N have 50% MinSOC should have at all times. The level is determined by the customer,
20% of N have 70% MinSOC and if MinSOC > SOC the Aggregator must charge the PHEY. And,
therefore, not available to provide control power. Customers tend to
choose mid alternatives when it comes to these kinds of choices [22).
f!..bat Battery capacity 11.3 kWh Battery capacity equivalent to the one in the Volvo V70 PHEV [23).
Phome Connection capacity home 3.5 kW Standard connections in German homes [18).
Pwork Connection capacity work 3.5 kW
Vwork Opportunity to charge at 80% Not likely that all companies can offer charging to their employees.
work Numerical value taken from [16). Note that all PHEVs can charge at
home.
SOCo Initial SOC SOCo E U(0.5, 1.0) Assumed to be rather high.
Thome Departing time from home Thome E N(08:00, 2.0) hours Departure times are assumed to be normally distributed throughout
Twork Departing time from work Twork E N(l7:00, 2.0) hours the aggregation. Collected from [24) and swiftly verified with [16).
Tdist Travel distance Mean value: 24 km, median: 8 km Log normally distributed for customers living in urban areas. Collected
from [24) and verified with [7) and [25).
Average speed 30 krnlh An assumed average speed in urban areas [6).
Electric energy consump­ 0.187 kWhlkm This number is derived from the fact that a typical mid size sedan will
tion when driving require this amount of energy for all electric operation [26).
'fJ Charging efficiency 94% [6).
wrso Secondary control power wrso E N(-67.0, 328.0) MW Derived from historical demand collected from the Amprion control
demand zone [27). The control zone is geographically located in the west of
Germany, and has an approximated secondary control demand of 1100
MW, symmetrically.
<I>Agg Bids on capacity and en- Mean values of historical accepted The actors' bids are collected from [28) for 9 consecutive months (July
ergy prices bids 2009 - March 2010).
¢Agg Bid size 4 MW symmetric Derived from the size of aggregation, connection capacity and ex­
pected availability.
O<fixed Fixed charging price 85.1 €IMWh 20% discount on the "Berlin klassik privatstrom" contract from 2008
[6).
f3bat Battery degeneration cost 30 €IMWh The cost of battery degeneration due to an extra cycle caused by
discharging the battery for control power. Only considered for part
Generation. Numerical value taken from [6).
"'bat The actual transfer capac- Linear if SOC < 80%. Follows a Approximated from a charging curve taken from [24).
ity logarithmic curve otherwise.

in set A is largest in the morning, due to continuous evening bid. The aggregation has 5000 PHEVs, which will demand an
and night charging. Many PHEVs are collected in D over the aggregation of approximately 100000 PHEVs to meet a 4 MW
day because the PHEV owners go to work, and not all have symmetric bid for part Controllable Load. Note that 1 MW of
charging opportunity there. Also, we can see that Bl and B2 offered capacity has an estimated worth of 7000 €/month,
are closely correlated. When the amount of PHEVs decreases which evidently leads to small potential income per PHEY. In
in Bb B2 decreases as well, i.e. PHEVs move from B2 to Bl addition, due to the large fluctuations in control power, the bid
to increase the down control capability. will be precisely met at the time 05:00, but will have a very
The subplots in Fig 6, follow the same principles as in Fig 4, large surplus at the time 20:00. Evidently, it is not cost efficient
but are the results from the simulations for part Controllable to let so many PHEVs share such a small capacity payment.
Load. As seen, neither the down nor the up control bid is We can directly conclude that part Controllable load offers
ever met. In fact, the Aggregator's control power capability less good control power capability than Generation, which
at the time 05:00 is almost zero. This is due to the fact that means that a discharging function is crucial when considering
the PHEVs in B2 will charge to be able to deliver up control symmetric bids.
power. However, up control is seldom activated by the TSO, In Table IV, we can see the mean income for a period of
which means that they will not stop charging. Inevitably, these one month for Generation and Controllable Load, respectively.
PHEVs will move to set A. To compensate the loss of up There will be low incomes for the PHEVs in the Basic Case.
control power when this happens, the Aggregator must move The largest total income are obtained from Generation, and
PHEVs in Bl to B2. Over time, this will result in a further loss are on the level of 17.9 €/monthIPHEV (the income for part
of both down and up control power. In addition, the frequent Controllable Load should be zero due to non-fulfilled bid). In
activation of down control power from the TSO, also moves comparison with the result of 81.7 €/monthIPHEV in [6], this
the PHEVs in Bl to A. The only action that can lower the result is small. However, [6] is an optimal study, assuming
SOC, and give a positive effect on the control power capability, that the Aggregator has the best bids, and that the bid is
is when the PHEVs are out driving. But, evidently, this is not precisely met at all times. Evidently, this is a part of an optimal
sufficient. approach. However, it is not realistic. Our result is reasonable
Further on, the circled curve at the time 05:00 must be because the bids are not perfectly matched at all times, and
lifted approximately 20 times to reach a level that matches the that they are not the best bids. Note that no costs for the
6

I ----e-- Down con vol power capability Up c ont o


r l pOW1!r capabihty

1
Total ability to deliver down control power
20,------,------,-- -1 -- Delivered down control power
50 so
-The down control bid
45 45
15
40 40

10 35 35

30 30

25 25

20 20

15 T 15 g
l EJ
10
time [hrsj
10
I ---e-- g
20,-- ----,------,- ---11
Total ability to deliver up control powe

- 0elivered up control power


9 J...
.
-The up control bid BasIc Cast ease 2 Casl! 3 Case 4 BasIc Case Case 2 Case 3 Case 4
15

Fig. 7. Box plots of the results of the sensitivity analysis for part Generation.

time [hrs]
stable towards changes in properties of the PHEV fleet for up
control. In Case 3, a significant increase of up control power
Fig. 6. Ability to deliver control power for part Controllable Load. can be observed. In addition, the relative difference between
the minimum and maximum level of up control power over
TABLE 11\ the day is rather small. Therefore, it is concluded that part
SHOWING THE PARAMETERS THAT WILL BE CHANGED IN THE
S ENSITIVITY ANALYS IS . Generation can be improved by a decrease in aggregation size,

Parameter New value Motivation Denoted as


combined with an increased connection capacity. However, the
N 15000 PHEVs Only studied for Control- Case 1 outcome is different when it comes to down control power
lable Load capability. Here, it can be observed that the fluctuations in
f:l.bat 25.0 kWh An aggregation with pure Case 2 control power are high for Case 3. The lowest point in
electric vehicles
aggregated control down power is lower than the bid, i.e. high
Phome, 11.1 kW Future scenario where fast Case 3
Pwork charging is more common charging availability and fast charging have a negative effect
lIwork 100% Increased implementation on the down control capability.
of public charging
<P Agg Prices on energy so activa- Case 4 In comparison to the Basic Case, an increased battery size
tion by the TSO is always and a change in bid strategy give positive effects on the down
valid control capability. Evidently, a change in bid strategy implies
that up control will be called more often, and, therefore,
increase the aggregated storage when the batteries discharge.
needed V2G infrastructure are included in the model, except As seen in Fig. 8, the results for part Controllable Load
for the cost for battery degeneration. This means that the result are different. Here, it can be seen that the up and down
can be lower in reality when these costs are included for the control power capabilities are very much alike. This is because
Aggregator. of the correlation between Bl and B2, i.e. the aim for the
Aggregator to distribute the PHEVs into set Bl and B2, so that
C. Sensitivity analysis equal up and down control power capabilities are maintained.
In this section, a sensitivity analysis is performed. The Furthermore, Case 1 gives the largest increase of control power
idea is to change some specific parameters each at a time, capability. However, the fluctuations are very big. Also, due
and see how the control capability and income changes for to the fact that the lowest point in control power over time
the Aggregator. The parameters chosen for this analysis can sets the size of the needed aggregation, the results of this
be found in Table III along with the new corresponding case are not optimal. Instead, Case 2 gives better results. The
values. The scenarios considered interesting are an increased total control power capability is increased in comparison to
(i) PHEV fleet size, (ii) battery size, (iii) connection capacity the Basic Case, and, in addition, a more stable control power
and availability, (iv) and a change in bid strategy. capability over the day is obtained. In comparison with Case
In Fig. 7, two box plots are displayed. These box plots 2, a doubling in battery size gives a higher lowest value in

show statistics for the up and down control control power control power, than a tripling in size of the PHEV fleet.
capabilities for the different cases over the simulated period of A surprising result at first sight is that an increase in connec­
24 hours for part Generation. The plots show the smallest and tion capacity gives a decrease in control power in comparison
largest values, the lower and upper quartiles, and the medians to the Basic Case. However, this is because located PHEVs in
for the control power capability. The results for the Basic Case B2 will only move faster to A, and will be compensated with
are also included as a reference case. PHEVs in B1. Therefore, both capabilities will decrease over
As we can see in the subplot to the right, the results for the time. Also, the results for Case 4 gives an identical result to
up control power capability is almost identical to the Basic the Basic Case. This is because the positive effect of providing
Case for all cases. This means that part Generation is close to up control by stopping the charging PHEVs in B2 are canceled
7

Down control power capability Up control power capability


Load. This is because charging opportunity is offered both at
I
I I
, I I home and at work for most of the PHEV owners. However, a
I I
I I
I I low average SOC gives a good effect on both the up and down
5 I I
I I control power capabilities. This is because the Aggregator can

�g �Q
4 distribute more PHEVs in the Bl and B2 sets, i.e., the PHEVs
3 can be distributed so that the up and down control power

I
T I ,- capabilities become more equal. Also, a low SOC increases
2 I
I I
I -,-
I
I ' T the actual transfer capacity, which results in more contributed
,8 I
g g 9

-'- 1 -
control power per PHEY. So, in order to reduce the average
-'- -'-
0 SOC, it could be wise to only offer charging opportunity at
BasIC Case Case 1 Case 2 Case 3 Case 4 a BaSIC Case Case 1 Case 2 Case 3 Case 4
home, and concentrate on providing control power for fewer
hours at, e.g. night. An additional advantage by delivering
Fig. 8. Box plots of the results of the sensitivity analysis for part Controllable
at night, is the high expected availability of PHEVs at these
Load.
hours.
TABLE IV
MEAN INCOME FOR ALL THE CASES IN €/MONTH/PHEV.

A. Bidding strategies

With the knowledge obtained from the simulations of the


Basic Case,part Generation could be improved by:
• Up control bids
• Fast charging
out by the frequent activation of down control by charging the • Fever PHEVs in the aggregation
PHEVs in B1.
The mean income for all cases can be observed in Table IV. Part Controllable Load could be improved by:
For part Generation the income is about the same for all cases, • Down control bids
except for Case 4. Here, the income is almost reduced by a • Slow charging only at home
factor two compared to the Basic Case, and this is due to • More PHEVs in the aggregation
the bid. The Aggregator puts the lowest energy prices for up • Bigger batteries
control, so that he always gets activated as soon as there is a With the bidding strategies presented here, it is predicted
control demand from the TSO. Delivering up control power that the fluctuations in control power capability would be min­
by discharging the PHEV is very expensive because: imized, and, therefore, the income per PHEV be maximized.
• Energy already bought and stored in the battery is dis­
charged
• Losses due to non perfect charge/discharge efficiencies B. Primary and tertiary control
• Costs for battery degeneration by increasing the amount
Although the simulations handle the secondary control
of cycles
market, a short discussion regarding the other two control
Evidently, this is not covered by the up control energy markets is presented here with the earlier results working as
price, and therefore, reduces the total income significantly. So, a foundation.
even though this bid strategy implies better symmetric control On the primary control market in Germany, the actor
power capabilities, the Aggregator cannot have this bid due to must put symmetrical bids with a delivery period of one
decreased income. month. From the results of the mathematical model showing
For part Controllable Load, Case 2 and Case 4 give an unfavorable down control power capabilities compared to up
increased income compared to the Basic Case. Case 2 is more control power capabilities, we do not recommend PHEVs to
profitable due to the fact that each PHEV can be active in the participate on this market. In addition, due to the fact that
bid a longer time because of the bigger batteries, and receive there are no energy payments and up control power is very
capacity payments accordingly. Case 4 gives a positive effect expensive to provide by the aggregation of PHEVs, this can
on the income due to the fact that up control power is delivered be really costly.
by just stopping the charging, i.e. the up control energy price The tertiary control market in Germany is asymmetric and
is a pure profit for the PHEV owner. has a short contract time. In addition, the bidding process
However, the incomes calculated for part Controllable Load is completed the day before delivery, and compensation is
should be taken cautiously. The bid is not fulfilled most of the obtained for both capacity and energy. All these properties
time, which in reality implies that no PHEV owner should be could be favorable for PHEVs. Flexible markets with short
compensated. time frames and specific delivery hours are preferable, where
the Aggregator can match the periods when the largest and
V. DISCUSSION most stable control power capability can be obtained from
Generally, there will be a high average SOC in the aggre­ the PHEVs. It is recommended to conduct more fine grained
gation over the day for both part Generation and Controllable studies of this market.
8

VI. CONCLUSIONS 2008.


[2] Vattenfall, http://www.vattenfall.se/sv/om-forsmark.htm. January 2010.
In this paper, we derived strategies for an Aggregator on [3] w. Kempton, J. Tomic, S. Letendre, A. Brooks, and T. Lipman, "Vehicle­
the German control markets. to-Grid Power: Battery, Hybrid, and Fuel Cell Vehicles as Resources for
First, we concluded that the Aggregator could provide Distributed Electric Power in California," University of Delaware et aI,
Tech. Rep.
symmetric control power by two different configurations, viz. [4] CEJoumal, http://www.cejoumal.netl?p=2024. May 2010.
part Generation and part Controllable Load. Here, part Con­ [5] W. Kempton and J. Tomic, "Vehicle-to-grid power Implementaion: From
stabilizing the grid to supporting large-scale renewable energy," Journal
trollable Load was a subset of part Generation, but without
of Power Sources, pp. 280-294, 2004.
a discharging function, i.e. only the charging of the PHEV [6] S.-L. Andersson, A. Elofsson, M. Galus, L. Goransson, S. Karlsson,
was controlled. Furthermore, for the Aggregator to assure the F. Johnsson, and G. Andersson, "Plug-in hybrid electric vehicles as
regulating power providers: Case studies of Sweden and Germany,"
TSO that he fulfills the symmetric control bid, we developed
Energy Policy, pp. 2751-2762, 2010.
a grouping process to accomplish this. The grouping process [7] C. Guille and G. Gross, "A conceptual framework for the vehicle-to-grid
placed the PHEVs into different sets with respect to, e.g. (V2G) implementation," Energy Policy, 2009.
[8] A. Brooks and T. Gage, "Integration of Electric Drive Vehicles with the
MinSOC, SOC and availability, and could therefore calculate
Electric Power Grid - a New Value Stream," EVS i8 Berlin, 2001.
the up and down control power capabilities. It was concluded [9] w. Kempton, V. Udo, and et al, "A Test of Vehicle-to-Grid (V2G)
that the Aggregator could control the distribution of PHEVs for Energy Storage and Frequency Regulation in the PJM System,"
University of Delaware et aI, Tech. Rep., 2008.
in the sets Bl and B2, to maintain equal up and down control
[10] w. Kempton and 1. Tomic, "Vehicle-to-grid fundamentals: Calculating
power capabilities. capacity and net revenue," Journal of Power Sources, 2004.
A collection of justifiable data and assumptions resulted [11] X. Zhong, A. Cmden, D. Infield, P. Holik, and S. Huang, "Assessment
of Vehicle to Grid Power as Power System Support," iEEE, 2009.
in the Basic Case that laid a foundation for Monte Carlo
[12] E. Larsen, D. Chandrashekhara, and J. Ostergard, "Electric Vehicles
simulations of a PHEV fleet used for commuting purposes on for Improved Operation of Power Systems with High Wind Power
the secondary control market. These simulations would vali­ Penetration," iEEE Energy2030, 2008.
[13] B. Williams and K. Kurani, "Commercializing light-duty plug-inlplug­
date the two configurations with the corresponding grouping
out hydrogen-fuel-cell vehicles: "Mobile Electricity technologies and
processes. opportunities," Journal of Power Sources, 2006.
The results from the simulations showed that part Gener­ [14] C. Quinn, D. Zimmerle, and T. Bradley, "The effect of communication
ation had very good up control power capabilities, but that architecture on the availability, reliability, and economics of plug-in
hybrid electric vehicle-to-grid ancillary services," Journal of Power
the down control power capabilities fluctuated more and had Sources, 2009.
a critical period in the morning. So, it was concluded that it [15] S. Han, S. H. Han, and K. Sezaki, "Design of An Optimal Aggregator
could be disadvantageous to provide symmetric control power for Vehicle-to-Grid Regulation Service," iEEE, 2010.
[16] 1. Fluhr, K.-H. Ahlert, and C. Weinhardt, "A Stochastic Model for
due to the differences in up and down control power capa­ Simulating the Availability of Electric Vehicles for Services to the Power
bilities. Part Controllable Load showed poor control power Grid," iEEE, 2010.
capabilities, and it was concluded that a discharging function [17] R. Follmer and U. Kunert, "Mobility in Germany (MiG) 2002 - Setting
standards for Travel surveys," i9th Dresden Conference of Traffic and
was crucial for symmetric bids. Transportation Sciences, 2002.
In order to maximize the reliability, efficiency and prof­ [18] S.-L. Andersson and A. Elofsson, "Plug-in Hybrid Electric Vehicles as
itability of the control power provided from PHEVs, it was Control Power," Master's thesis, Chalmers Technical University, 2009.
[19] N. Ingvar and C. Persson, "Vehicle to Grid State of the Art System
shown that part Generation (start/stop charging/discharging) Design - Plug In Hybrid Electric Vehicles as source of power for
only should make up control bids. Furthermore, it was shown the control markets in Sweden and Germany," Master's thesis, Royal
that part Controllable Load (start/stop charging), should aim at Institute of Technology, Apr. 2010.
[20] G. Schulz, Phone interview and mail correspondence, Vattenfall Europe
down control bids during limited time periods, e.g. at night. Transmission, 2009.
By doing so, the fluctuations would be minimized, i.e. the [21] M. Dupuy, "Electricity Markets Balancing Mechanisms and Congestion
difference between the lowest and highest value in control Management," Master's thesis, Royal Institute of Technology, 2008.
[22] I. Simonson, "Choice Based on Reasons: The Case of Attraction and
power would be reduced, and therefore the income per PHEV Compromise Effects," Journal of Consumer Research, vol. 16, 1989.
would be maximized. [23] Hybridmile, http://www.hybridmile.comlnewslvolvo-vattenfall-enerI/.
Note that the results from the simulations are only valid for June 2010.
[24] M. Rouselle, "Impact of Electric Vehicle on the Electric System,"
the secondary control market. Actual simulations performed Master's thesis, Royal Institute of Technology, 2009.
with respect to the constraints of the primary and tertiary mar­ [25] S. Bergman, "Plug-in hybrider elhybridsfordon fOr framtiden;' Elforsk,
kets may give results that differ from those of the qualitative Tech. Rep., 2008.
[26] T. Markel and A. Simpson, "Plug-In Hybrid Electric Vehicle Energy
discussion in section V. Storage System Design," NREL, 2006.
[27] Amprion Gmbh, http://www.amprion.netlen. January 2010.
REFERENCES
[28] Regelleistung, http://www.regelleistung.net. January 2010.
[1] D. Salmonsson and V. Neimane, "Plug-in Hybrid Electric Vehicles:
Capacity Study in European Perspective," Vattenfall R&D, Tech. Rep.,

You might also like